I have looked at http://apache-ignite-users.70518.x6.nabble.com/Ignite-Write-Behind-performance-td5385.html which seems similar, but does not answer my question.
I am using write behind caching with a single server it has 16 cores and 32 G of memory. My write behind config is setWriteBehindEnabled (true) setWriteBehindFlushSize(1024 * 100) setWriteBehindFlushFrequency(5000L) setWriteBehindBatchSize(512) setWriteBehindFlushThreadCount(1) My intent with the config is to have plenty of room in the buffer, do large batches to the database, but never fall more then 5 seconds behind if things are not busy. I have a main thread that reads from kafka and writes the cache. Depending on the message, different caches may be written, but they all use the same settings in the cache config documented above. What I am seeing is a degradation in performance on the main thread as the database fills and the writes on the background threads take more time. The write behind buffer is not filling, I am not seeing cache critical size errors. Other than contention for the system resources, is there any reason that I should see a slow down on my main thread which does the cache put? Is there any locking on writing of the grid that happens when the background threads pull from the write behind buffer? -- Sent from: http://apache-ignite-users.70518.x6.nabble.com/
